Most products have no idea what their AI agents did yesterday

## Executive Summary Velt (YC W22) is evolving from a specialized collaboration SDK provider into a critical infrastructure layer for the "Agentic Era." By providing a unified, immutable activity log that tracks both human and AI agent actions, Velt addresses the growing accountability and auditability gap in software. This is a timely "picks and shovels" play that capitalizes on the transition from AI chatbots to autonomous AI collaborators. ## Founder / Team Assessment The team is YC W22 alumni, indicating a high baseline for technical execution and product-market fit (PMF) iteration. Their ability to pivot from general collaboration tools (comments, CRDTs) to a specific, high-pain-point AI infrastructure problem (agent accountability) demonstrates strong market intuition. However, the team size and specific technical depth in distributed systems/security—crucial for "immutable" enterprise logs—remain unverified and require further investigation during diligence. ## Market Analysis The market for AI Observability and Governance is expanding rapidly as enterprises move AI agents from experimental sandboxes to production environments. While the collaboration software market is large ($20B+), Velt is effectively targeting the "AI Trust & Safety" and "Compliance" verticals. As regulatory frameworks (like the EU AI Act) begin to mandate audit trails for automated decisions, the demand for standardized, third-party logging infrastructure will shift from "nice-to-have" to a mandatory enterprise requirement. ## Product / Traction Velt’s core product is a "collaboration-in-a-box" SDK, which provides a high-moat integration point within a customer's codebase. The new Activity Logs feature leverages their existing real-time infrastructure (CRDTs, presence) to offer a unified view of "who did what," regardless of whether the actor is a human or a bot. Traction signals are moderate; while their YC pedigree and recent "Show HN" engagement show developer mindshare, we lack hard data on enterprise seat counts or API call volumes. ## Competitive Landscape Velt faces competition from two sides: 1. **AI Observability Startups:** Companies like LangSmith or Arize focus on LLM performance (latency, cost, traces), but often lack the UI-level context of what the agent actually *did* to the document or data. 2. **Incumbent Infrastructure:** Auth0 (Okta) or Datadog could expand into agent-specific auditing. Velt’s advantage is its position at the **application layer**. By being the SDK that powers the comments and edits, they are uniquely positioned to log the *intent and outcome* of an action, not just the raw API trace. ## Investment Thesis **Bull Case:** 1. **The Standard for Agent Identity:** Velt becomes the "Okta for Agents," providing the definitive record of agentic behavior across SaaS platforms. 2. **High Switching Costs:** Once an SDK is integrated into a product's core collaboration workflow, it is rarely ripped out, leading to high LTV and negative churn. 3. **Regulatory Tailwinds:** Global AI regulations will soon mandate the exact "immutable audit trail" Velt has already built, forcing rapid enterprise adoption. **Bear Case:** 1. **Feature vs. Product:** There is a risk that "Activity Logs" are viewed as a feature that internal engineering teams will simply build themselves using standard database triggers. 2. **Market Crowding:** The AI infrastructure space is currently overcapitalized; Velt may struggle to maintain pricing power against open-source alternatives. 3. **Integration Friction:** SDK-based sales cycles can be long and require significant developer "buy-in" compared to simple API-based observability tools. ## Recommended Action **Conduct Deeper Diligence.** We need to validate the "stickiness" of their existing collaboration SDK and determine if the Activity Log feature is driving new enterprise pilots or merely serving as a retention tool for existing YC-scale customers.